- Fix default layout with TensorFlow models (#34)
- Support for QuPath v0.5.0
- Improved support for TensorFlow via Deep Java Library
- Optionally use the model name for a model stored in the user directory
- May be in a 'stardist' or 'models' subdirectory
- Provides an alternative to specifying the full model path
- Reduce non-essential logging messages to 'debug' level
- Support for QuPath v0.4.0 and Deep Java Library (via the Deep Java Library extension)
- Support for initializing the model builder from a Bioimage Model Zoo spec
- Support for preprocessing based upon the full (downsampled) image (#20)
- Support for both per-channel and joint channel normalization (#14)
- The documentation is now here
- Fix bug applying detection to a ROI on a z-stack / time series, whereby the default plane would be used for detection (at least if padding > 0).
- Invalid operation for scaling (#17)
- QuPath with CUDA doesn’t release GPU memory after StarDist segmentation Usage & Issues qupath (#11)
- You'll need to add
stardist.close()at the end of any scripts (assuming you've usedstardistas the variable name, as it is here)
- You'll need to add
- Reduces some occurrences of the dreaded
TopologyException- More will hopefully be removed in the next main QuPath release
- First version, written to be compatible with QuPath v0.3.0